AC heaters ultimate hum limit

[ruffrecords]

I have made the twisted wire heaters modification. The 50Hz hum looks to be 6dB lower but the 100Hz looks within a dB of what it was. As requested by Ricardo I have captured pics with various numbers of points in the FFts. Each is labelled with the number. As Ricardo suggested, the broadband noise floor gets lower with more points but the 50/100Hz component amplitudes remain unchanged as does the total power

Thanks for this Ian.  This tells me the QA400 software is "calibrated" to read single tones 'properly'.

What you have to avoid is to quote the 'FFT noise floor' as the noise level ... this is almost always dominated by the FFT size. 

Some measurement pseudo gurus, including the authors of a couple of popular measurement packages are guilty of this very misleading sin.

The proper figure to quote is the 'total power' level over a given bandwidth.  This is also what you should use if you want to say "hum is 20dB below the noise".

Yes I am well aware of this. I have seen some audiophile preamp articles with spectra and the authors claiming the noise level is the same as the baseline noise of the spectra. I am still confused how the number of points affects this. However, if the spectra relatively flat and is shown as V per root Hz then you just need to add 20 log( root(20KHz)) = 43dB to get the total.

Cheers

Ian

 

[ruffrecords]

I started using capacitance multipliers for HT because they need no heatsinks and work well enough for me. Mains voltage changes are small enough, they also make tube stages very quiet. Smallest component for my typical multiplier is TIP50 transistor, the rest are 2 5W resistors, about 100u/400V input filter cap, 3 33u/400V and a few 5W zeners into TIP50 base. Very simple, almost no heat, beside 100u cap others could probably be smaller than 33u and film types. This can be easily built p2p on veroboard or so.
Can't find examples where people use cap multipliers for heaters, some audiophiles put them before the regulators to get lower ripple.
Wouldn't be this good solution even at lower voltages/higher currents for heaters? I found out input filter for HT needs bigger lytic cap like 100u at ~250V/15mA, others like cap into transistor base can be even smaller than 33u even at beta of 50, can't hear or see any problems on a scope this way. Fearn uses only simple CRC filters with zeners and no one complains, that is even simpler.
Material is so cheap that every channel in larger systems could have their own cap multipliers, feeding them with AC shouldn't be a problem. I'm sure about HT, although never tried them for heaters because i got stash of heatsinks for free and never get over 1A at 6,3V.  Would capacitance multiplier be problematic for heaters, is there any reason for it?

The very early incarnations of the EZTubeMixer PCBs had an individual 7812 regulator on board with roughly smoothed dc being fed in via the backplane. But as the design progressed there was no longer any room for the regulators so I moved to an external regulated dc heater supply. At one point I briefly played around with fitting the 7812s on the backplane PCB by each channel but again there was not really enough room. When I started the LunchBox project I was looking for ways to save cost and tried ac heaters. To my surprise the noise level measured by my Lindos test set was unchanged from the dc case so I started using ac heaters. It was not until I recently tried using the QA400 that it became apaprent how much hum was actually there.

Cheers

ian

 

[kambo]

why DC heater supply needs to be regulated ?
isnt it as good as regulated using one of the design from Cj' post
no need for heat sink! just couple of diodes and caps!

http://groupdiy.com/index.php?topic=57166.0

electricity is pretty stable lately!

 

[trobbins]

Ian,
Your pcb layout doesn't really even try to reduce hum pickup from capacitive coupling to the second triode - pin2 grid on each tube couldn't really be any closer to the heater!!!  And you are using 12VAC, which makes capacitive coupling even worse, and any humdinger pot is not really going to be able to achieve any sort of a null.

Perhaps if you change to 6V3 powering, and use a twisted screened data cable to get the heater to each group of pins 4 and five, and keep the cable 'complete' to a point as close to the terminals as possible.  The cable shield should terminate on a pad in the middle of the pins, as per a socket spigot, which should be your ground plane 0V.  It looks like you are using a ground plane on the other side?

What heater elevation are you using, and what is the cathode voltage of the follower?

Ciao, Tim

 

[ruffrecords]

why DC heater supply needs to be regulated ?
isnt it as good as regulated using one of the design from Cj' post
no need for heat sink! just couple of diodes and caps!

http://groupdiy.com/index.php?topic=57166.0

electricity is pretty stable lately!

Smoothed but unregulated dc is fine if you know exactly how many tubes you will be powering.  The EZTubeMixer designs are modular so the number plugged in is variable.

Cheers

Ian

 

[ruffrecords]

Ian,
Your pcb layout doesn't really even try to reduce hum pickup from capacitive coupling to the second triode - pin2 grid on each tube couldn't really be any closer to the heater!!!  And you are using 12VAC, which makes capacitive coupling even worse, and any humdinger pot is not really going to be able to achieve any sort of a null.

Agreed; it was designed for dc heaters. As I mentioned earlier I only tried ac heaters in an attempt to simplify power supplies and reduce costs for certain applications.

Perhaps if you change to 6V3 powering, and use a twisted screened data cable to get the heater to each group of pins 4 and five, and keep the cable 'complete' to a point as close to the terminals as possible.  The cable shield should terminate on a pad in the middle of the pins, as per a socket spigot, which should be your ground plane 0V.  It looks like you are using a ground plane on the other side?

The classic solo version of this design has a 6.3VAC heater option. The current prototype uses 12VAC but I have just taken delivery of some custom mains toroids with a 6.3VAC heater winding. I need to test these so I will do a before and after comparison of hum in the prototype.

What heater elevation are you using, and what is the cathode voltage of the follower?


Ciao, Tim

Heater elevation is to 25% of the HT supply voltage. Top cathode of mu follower is close to 50% of supply. Supply is nominal 300V so the botom triode's cathode sees about +75V Vhk and the top one about -75V.

Cheers

Ian

 

[kambo]

regarding your QA400:

http://groupdiy.com/index.php?topic=51698.msg757483#msg757483

EDIT:

I think balanced ins and outs are a good idea.  My main problem with this unit is 60Hz hum pickup, which seems to mostly be picked up by the BNC cables to and from the qa400 (as moving them around quite noticeably changes the amount of hum).

 

[ruffrecords]

I have created a trial PCB layout of a Lo Hum version of the Classic Solo PCB. I have rotated to two tubes so their heater pins are close to each other and strapped them directly together (the two red vertical traces in the attached image). I would add a pad half way along each trace and connect the AC directly there via a twisted pair. I have manages to keep the grid traces well away form the heaters and the only tracks that cross the heaters are the cathode output of the first stage and the anode trace of the second. The only thing I don't like is the right hand thick red trace which is the HT between the two tubes as is cuts off the ground plane to the centre of the tubes and the one side of the heaters (the heaters are on the same side as the ground plane).

Comments?

Cheers

Ian

 

[JohnRoberts]

Maybe make your own star-quad version of twisted wires using 4 wires (2 red and 2 black) interleaved for a more symmetrical wrap.  The 2 wire twisted pair is often a little lumpy and not very symmetrical. IIRC the star quad uses a cotton filler in the very middle of the wire wrap to keep it symmetrical.

I can think up things for other people to try all day long.  ;D

JR

 

[kambo]

i would move C1 to the left side of the board, and place C8 between two tubes (cross_over heater traces, looks like long enough component).  i dont think 1/2 inch more gap between tubes gonna make any difference for ur heater supply hum...

edit: innertube audio guys are using octal based tubes where possible... they think tubes with big solid bottom, sounding better, in your case, i would give it a try using tube sockets, it may improve hum !

 

[moamps]

Comments?

I would do it this way (more or less):

 

[ruffrecords]

Many thanks for the layout ideas. Plenty of food for thought.

Cheers

Ian

 

[Gene Pink]

Maybe make your own star-quad version of twisted wires using 4 wires (2 red and 2 black) interleaved for a more symmetrical wrap.  The 2 wire twisted pair is often a little lumpy and not very symmetrical. IIRC the star quad uses a cotton filler in the very middle of the wire wrap to keep it symmetrical.

Or perhaps just use star quad. Website says it is 24awg X 4, So effective 21awg there and back, about 30 m-ohms per running foot. Too thin? Too ugly?

I can think up things for other people to try all day long.  ;D

John, if you think something up, it is gonna be worth trying.

Gene

 

[trobbins]

Ian,
The pcb layout shows 240VAC input. 

Although it is pedantic, and unlikely to show any measurable change, it is worthwhile trying to make the trace layout as low noise as possible. 

For example, BR1+ should go direct to C14, and closely spaced from the BR1- trace.  There should then be a seperate trace from C14 pos to R21. 

Similar concept with traces in and out of C15+ pad, and from C16+ pad.

Similar for 48V phantom.  Reg input should go to C13+ pad.

Any part that is meant to bypass high frequency transients, such as C18, should have its traces presenting a minimum loop area - so the 'long' trace should preferably pass underneath the cap, and then track around to BR1  with as little distance as practical.

Hope that helps.

 

[ricardo]

You can, but you must use the PSD and not the FFT, it means that your Y axis scale will be V²/Hz or V/sqrt(Hz), ..  and other good stuff

But then the Y axis/scale is 'wrong' / misleading for the hum/buzz spikes.  You see this a lot even from some of the people who write measurement software and should really know about such gotchas.

As I said, the most common symptom of such naivety is the "hum/buzz components are x dB above/below the noise floor" claims.

[quote author=ruffrecords]It was not until I recently tried using the QA400 that it became apaprent how much hum was actually there.[/quote]
From what you said about the LINDOS measurements, your hum appears to be well below the audible 'white' noise level.

It's just that a big FFT measurement reduces the "noise floor" so emphasizing any monotone elements like hum/buzz.

One way to look at it is that an FFT takes the total wideband (white/pink bla bla) noise and divides it up into into its various frequency 'bins',  Larger FFTs have more bins so there's less noise in each individual bin.

A single tone will sit only in one bin so will stand up further & further as the FFT size gets bigger.  All this happens with swept analyser instruments like a B&K 2010 but its more obvious with FFT gear that lets you change the FFT size easily.

All this is good if what you are looking at is usually swamped by white noise .. eg signals from inter-planetry space probes but here you might be doing a lot of work on something that has no audible or (conventionally) measured advantage.

 

[ruffrecords]

Ian,
The pcb layout shows 240VAC input. 

Although it is pedantic, and unlikely to show any measurable change, it is worthwhile trying to make the trace layout as low noise as possible. 

For example, BR1+ should go direct to C14, and closely spaced from the BR1- trace.  There should then be a separate trace from C14 pos to R21. 

snip

I could swap the positions of J8/BR1 ans C18 so the bridge is in line with C14. In other designs I usually ensure C18 is symmetrical about J8 to minimise trace length/area but I did not have room to do this. Your other points are easy to implement and I will remember them in future. Many thanks for the input.

Cheers

Ian

 

[ruffrecords]

[quote author=ruffrecords]It was not until I recently tried using the QA400 that it became apparent how much hum was actually there.

From what you said about the LINDOS measurements, your hum appears to be well below the audible 'white' noise level.
[/quote]

It is. In normal use even with the gain wound up to 70dB the hum is swamped by acoustic rumble. If you switch in a HPF to reduce the rumble it drops the hum too.

I don't need to reduce the hum but I would like to understand its causes and ways to minimise it if for no other reason than my own satisfaction.

Cheers

Ian

 

[ruffrecords]

Here is a tweaked version of the solo layout with the suggested PSU mods incorporated. Next I will look at the amps.

Cheers

Ian

 

[My3gger]

I started using capacitance multipliers for HT because they need no heatsinks and work well enough for me. Mains voltage changes are small enough, they also make tube stages very quiet. Smallest component for my typical multiplier is TIP50 transistor, the rest are 2 5W resistors, about 100u/400V input filter cap, 3 33u/400V and a few 5W zeners into TIP50 base. Very simple, almost no heat, beside 100u cap others could probably be smaller than 33u and film types. This can be easily built p2p on veroboard or so.
Can't find examples where people use cap multipliers for heaters, some audiophiles put them before the regulators to get lower ripple.
Wouldn't be this good solution even at lower voltages/higher currents for heaters? I found out input filter for HT needs bigger lytic cap like 100u at ~250V/15mA, others like cap into transistor base can be even smaller than 33u even at beta of 50, can't hear or see any problems on a scope this way. Fearn uses only simple CRC filters with zeners and no one complains, that is even simpler.
Material is so cheap that every channel in larger systems could have their own cap multipliers, feeding them with AC shouldn't be a problem. I'm sure about HT, although never tried them for heaters because i got stash of heatsinks for free and never get over 1A at 6,3V.  Would capacitance multiplier be problematic for heaters, is there any reason for it?

The very early incarnations of the EZTubeMixer PCBs had an individual 7812 regulator on board with roughly smoothed dc being fed in via the backplane. But as the design progressed there was no longer any room for the regulators so I moved to an external regulated dc heater supply. At one point I briefly played around with fitting the 7812s on the backplane PCB by each channel but again there was not really enough room. When I started the LunchBox project I was looking for ways to save cost and tried ac heaters. To my surprise the noise level measured by my Lindos test set was unchanged from the dc case so I started using ac heaters. It was not until I recently tried using the QA400 that it became apaprent how much hum was actually there.

Cheers

ian

So QA400 shows better picture of hum compared to Lindos. I didn't do such tests, only listened how quiet finished preamp is under different conditions and of course looked with a scope if there was anything wrong.
The last time i used ~ 60x80mm veroboard with multiplier for HT (276V/18mA) and LDO for heaters (6,3V/400mA). Zeners into TIP50 beneath veroboard, 5W resistors are barely warm and could go there too, beside nice 100u Epcos i could choose smaller size and still good caps for the rest, so board could be much, much smaller, without heatsinks. Phantom has TIP121 for it, although your TL783 solution would a lot simpler and smaller for such build.
I checked heat dissipation, noise reduction, ripple shape at the output, checked requirments for space; at least HT could benefit from C multipliers. Seems like heaters too, have to calculate this part too, examples i could find are only used as pre or post regulators for DHTs. That is for large systems like yours, there should be advantages. Do you think so?

 

[JohnRoberts]

I can think up things for other people to try all day long.  ;D

John, if you think something up, it is gonna be worth trying.

Gene

Back in the day when I managed several employees, my day job was actually keeping them busy, doing productive stuff... I have many ideas that aren't worth doing. I only pursue a small fraction of my own ideas myself, but thanks for the sentiment.

JR